Hi all,

Just released my book:

lgisp-cover1200×1548 72.3 KB

Backstory: after shipping my first customer-facing LangGraph agent for an enterprise security product, I realized how much of the framework I’d never actually touched. So I started writing small runnable examples to poke at each feature, with notes on what surprised me. The examples kept connecting to each other and somewhere along the way it became 31 chapters.

What’s in it:
• Part I is LangGraph on its own, one small piece at a time: state, nodes, edges, loops, checkpoints, interrupts. No LLM yet, on purpose.
• Part II brings in a real model: tool calling, structured output, MCP, memory, streaming, middleware, testing.
• Part III is the production stuff: serving, async, observability, RAG, evaluation, blue-green deploys, guardrails.

Things it’s done a little different from other books:

1. Every transcript in the book is a real run, pasted as-is, including the times the model did something dumb. And it all runs on a local model by default (Qwen3.8-27B via LM Studio), so you can follow along without an API bill.
2. Every chapter ends with Bronze/Silver/Gold challenges, worked solutions in the repo: GitHub - LgISP/langgraph-in-small-pieces: Code Repo for LangGraph in Small Pieces · GitHub
3. It has lots of silly illustrations.

Pinned to langgraph 1.2.14 / langchain-core 1.6.9.

It’s a beta. Every paragraph on the site has a comment button, and there’s an envelope icon in the sidebar that goes straight to me. If something doesn’t run for you, or a chapter answers the wrong question, I want to know. Those are the comments I’ll actually act on.

Hope you like it. Please check it out it’s free online at:

hitl1200×1080 44.4 KB

observe1200×1163 32.7 KB

loop-until1200×1129 34.3 KB